AgentMark

A declarative, extensible, and composable approach for creating LLM prompts using Markdown and JSX.

Overview

AgentMark is a Visual Studio Code extension that brings a new, powerful way to create language model (LLM) prompts. Designed with a focus on readability, portability, and syntax highlighting, AgentMark allows you to write prompts in a language-agnostic, declarative format using .prompt.mdx files.

By leveraging Markdown and JSX, AgentMark provides a clean, composable, and extensible solution to enhance your prompt development workflow.

Features

• Language Agnostic: Write prompts for any language model without platform or syntax constraints.
• Composable: Easily modularize and reuse prompt components across different files and projects.
• Readable: Promotes a declarative structure that enhances readability and maintainability.
• Syntax Highlighting: Enjoy rich syntax highlighting to improve development experience in .prompt.mdx files.
• Portable: Share and transfer prompts seamlessly across different environments.
• Type Safety: TS-style Type Safety out of the box

Running the Extension

AgentMark supports multiple types of generation:

Text Generation

---
name: text
text_config:
  model_name: gpt-4o-mini
---

<User>What's 2 + 2?</User>

Object Generation

---
name: example
object_config:
  model_name: gpt-4
  schema:
    type: object
    properties:
      event:
        type: object
        properties:
          name: 
            type: string
            description: The name of the event
          date:
            type: string
            description: The date of the event
          attendees:
            type: array
            items:
              type: object
              properties:
                name:
                  type: string
                  description: The name of the attendee
                role:
                  type: string
                  description: The role of the attendee
              required:
                - name
                - role
        required: 
          - name
          - date
          - attendees
---

<System>You are a math tutor that can perform calculations.</System>
<User>What's 235 * 18?</User>

Image Generation

---
name: image
image_config:
  model_name: dall-e-3
  num_images: 1
  size: 1024x1024
  aspect_ratio: 1:1
  seed: 12345
---

<ImagePrompt>
A futuristic cityscape at sunset with flying cars and neon lights
</ImagePrompt>

Speech Generation

---
name: speech
speech_config:
  model_name: tts-1-hd
  voice: "nova"
  speed: 1.0
  output_format: "mp3"
---

<System>
Please read this text aloud.
</System>

<SpeechPrompt>
This is a test for the speech prompt to be spoken aloud.
</SpeechPrompt>

1. Open Visual Studio Code.
2. Navigate to your .prompt.mdx prompt
3. Copy this, or create your own.
4. Press F5 or click the "Run" button in VS Code to launch the extension.

Modules

You can import .md or .mdx file within your files.

---
name: basic-prompt
text_config:
  model_name: gpt-4o-mini
---

import OutputInstructions from './output-format.mdx';

<User>
  What's 2 + 2?

  <OutputInstructions />
</User>

Props

Props can be accessed using {props.varName}. You can test props in your file config through the test_settings.

---
name: basic-prompt
text_config:
  model_name: gpt-4o-mini
test_settings:
  props:
    num: 3
---

<User>
  What's 2 + {props.num}?
</User>

Documentation

Comprehensive documentation, including guides and API references, is available in the AgentMark GitHub repository. Refer to the documentation for detailed instructions on using AgentMark features and integrating it with your workflow.

Community

Chat with our growing, tight-knit community. Join our Discord

Feedback

We value your feedback to continuously improve AgentMark. Please submit any issues, feature requests, or other feedback through the GitHub repository's issue tracker.